**** BEGIN LOGGING AT Tue Mar 28 03:00:04 2017 Mar 28 03:16:34 Sorry, I should have mentioned that I'm running "bone-debian-8.7-seeed-iot-armhf-2017-03-26-4gb" image atm Mar 28 04:26:04 hi guys. I just got a BB Blue and was trying to enable some dts overlays but the slots file is always missing. Is there a new way to enable runtime dts overlays for kernel 4.4-ti? Mar 28 04:34:38 Epilog: slots file has moved Mar 28 04:35:42 see http://elinux.org/Beagleboard:BeagleBone_Debian_Image_Migration Mar 28 04:45:37 Thanks zmatt. Unfortunately I get a "No such file or directory" whenever I run "sudo cat /sys/devices/platform/bone_capemgr" Mar 28 04:49:11 it's the directory *containing* the slots file, or at least it should be Mar 28 05:03:35 is there a special build for the Blue or the normal Debian img will do? Mar 28 05:04:17 was the slots file in the indicated place? Mar 28 05:05:55 I'd presume the blue simply comes with a jessie iot image, but I don't have a blue so I can't verify that Mar 28 05:06:39 that's the thing. I did a find /sys | grep slot and |grep cape and nothing apparent shows up Mar 28 05:07:14 can you copy/paste the content of your /boot/uEnv.txt into a pastebin ? Mar 28 05:07:46 (if the blue has internet you can probably do: pastebinit I wonder what they shipped onto those things Mar 28 05:09:32 I have a suspicion, but I just want to verify it :P Mar 28 05:10:40 one sec, I think I screwed up the boot. Let me make it boot again Mar 28 05:18:55 I'm using 'bone-debian-8.7-iot-armhf-2017-03-19-4gb.img' Mar 28 05:27:56 zmatt: ok done, what file did you want to see? Mar 28 05:28:59 zmatt: /boot/uBoot.env -> http://paste.debian.net/924682/ Mar 28 05:41:48 ew, u-boot overlays... Mar 28 05:48:32 zmatt: what does that mean? Mar 28 05:54:30 hi Mar 28 05:58:23 I wonder why this problem sometime block on line 44 https://pastebin.com/FVF24np8 Mar 28 05:58:36 without any error Mar 28 05:59:02 Epilog: it doesn't seem to be enabled by default though Mar 28 05:59:49 Epilog: u-boot overlays means that a base dtb and one or more overlays will be merged by u-boot into a single device tree that's passed to the kernel Mar 28 06:01:12 i.e. it retains all the ugliness of overlays, but removes its sole redeeming feature: dynamic loading. as a bonus, it makes the bootloader even *more* bloated Mar 28 06:01:23 why this is considered a good idea by some eludes me right now Mar 28 06:03:49 Epilog: sorry, I have no explanation of why capemgr seems to be missing for you, maybe try the mailing list? Mar 28 06:13:09 so in that case, should I alter anything in the uEnv.txt file? Mar 28 06:13:58 if you want to load overlays you probably don't want cape_universal enabled, but that doesn't solve the mystery of the missing bone_capemgr Mar 28 06:14:53 I saw that new dts were added for BB Blue and 4.4 in the github repo Mar 28 06:15:48 of course Mar 28 06:41:27 how would you get into a situation wher ethe DSB function is relevant Mar 28 06:41:32 err DSB instruction* Mar 28 06:42:08 i mean isn't everything already in order Mar 28 07:04:30 ayjay_t: dsb on an am335x does two things: Mar 28 07:06:01 1. it ensures all non-strongly-ordered writes have either arrived in cache (when applicable), its destination (when this is within the cortex-a8 subsystem), or the async fifo to the L3 interconnect Mar 28 07:06:31 2. it ensures all previously "executed" cache maintenance operations have completed Mar 28 07:07:56 a nice example is if you write executable code to memory, then before being allowed to execute it you need to: Mar 28 07:08:43 1. if the L1 cache policy is write-back: flush dcache to point of unification Mar 28 07:08:56 2. dsb Mar 28 07:09:11 3. invalidate icache Mar 28 07:09:37 sorry, "flush" should be "clean" of course Mar 28 07:11:27 if the L1 cache policy is write-back then the dsb is needed to ensure the dcache clean is done before the icache invalidate, while if the L1 cache policy is write-through the dsb is needed to ensure the previous memory writes themselves have at least hit L2 (which is the POU) Mar 28 07:16:14 I wonder why this problem sometime block on line 44 https://pastebin.com/FVF24np8 Mar 28 07:20:33 jamesaxl: yes, so you've said several times already, but I don't know exactly you expect from us... line 44 is a print statement Mar 28 07:20:41 *what you expect Mar 28 07:22:13 zmatt, when i loop the function ulta_sonic, it blocks without any error Mar 28 07:22:16 all file I/O (including to tty, pipe, or socket) can block temporarily or in some cases indefinitely Mar 28 07:23:06 I don't understand you Mar 28 07:23:39 ayjay_t: having fun with ARMv7 system-level architecture? :) Mar 28 07:24:52 zmatt, I have an ultrasonic and servo motor, I move my servo motor using potentionmeter and on every step, the ultrasonic detect if the distance is 10 cm Mar 28 07:25:14 yes this code looks very unreliable for that purpose Mar 28 07:25:35 everyhing works but when I put my hand long time on the ultasonic the problem blocks Mar 28 07:26:37 you mean it gets stuck in the loop on lines 31-32 because it has missed the pulse Mar 28 07:27:53 the am335x has peripherals that can accurately generate a pulse and measure when the reflection pulse arrives, with 10ns precision Mar 28 07:29:20 zmatt, which pulse are you talking about? Mar 28 07:30:18 in the last two lines I've referred to both of them Mar 28 07:32:42 * zmatt has evidently ran out of helpful Mar 28 07:35:07 zmatt, no need to run Mar 28 12:11:07 zmatt: nTRST resets the debug logic? that sounds promising. Mar 28 12:11:39 maybe Mar 28 12:11:49 quite unlikely actually Mar 28 12:12:49 tracing also works without jtag, and CTI's reset domain is that of the trace logic Mar 28 12:14:08 thinkfat: I pulled some strings for you, just got a reply "I asked one of our engineers from the JTAG/DebugSS team to please comment" Mar 28 12:14:13 :) Mar 28 12:16:21 zmatt: :-) Mar 28 12:16:29 zmatt: thanks, very much appreciated Mar 28 12:17:09 zmatt: I will nevertheless test the nTRST, trivial to check Mar 28 12:19:49 indeed Mar 28 12:20:25 zmatt: by the way, do you know if wakeup from a DS state with MOSC off is possible through UART0 or will I have to use the IO wakeup capabilites of GPIO pads? Mar 28 12:20:48 should be possible if the uart is setup correctly Mar 28 12:20:52 iirc Mar 28 12:21:45 well I know it wakes up from "standby", if I enable wakeup in the hwmod and also set the SIDLEMODE accordingly Mar 28 12:22:20 but in DS1 state, the system doesn't wake up again Mar 28 12:23:19 the main difference between standby and ds1 beeing MOSC on vs off Mar 28 12:25:46 so, I think you may need to explicitly enter the uart into sleep mode, in which case uart stops its internals clocks anyway hence there's no reason for it to require a clock to be supplied Mar 28 12:26:50 and then probably set RX_CTS_WU_EN, and relevant bit in UART_WER, and ENAWAKEUP in sysc Mar 28 12:26:56 something like that Mar 28 12:27:53 maybe WER and ENAWAKEUP suffice Mar 28 12:35:45 ENAWAKEUP I already do, but I might be missing the UART_WER Mar 28 12:36:10 zmatt, which beaglebone board do you have? Mar 28 12:36:44 mostly blacks (rev C, one A5A), and two BBEs Mar 28 12:42:44 zmatt, that is nice, did you do projects before? Mar 28 12:43:51 ? Mar 28 12:44:45 zmatt, i meant did you make something with these boards Mar 28 12:46:01 we use them at work, they get integrated into products Mar 28 12:49:27 zmatt, have you ever written a C++ code for, if yes, do you use a c/c++ lib or you just changing te content of pin file. Mar 28 12:50:42 a lib for.. what? changing the content of a file? Mar 28 12:51:23 although a sysfs attribute doesn't behave quite like a file, it doesn't really need a library either Mar 28 12:51:37 zmatt: CTI is completely unimpressed by a TRST Mar 28 12:52:02 zmatt: but I remember there is a configuration in ICEPick whether a TRST actually resets the debug logic or not Mar 28 12:52:05 will check Mar 28 12:56:06 zmatt, you know that we can change the value of gpio or getting the value of an Input Gpio in shell, Mar 28 13:00:35 I think I will write a C++ library Mar 28 13:38:40 zmatt: eh, TRST of course resets icepick, which means DBGEN gets released, which is exactly _not_ what I want Mar 28 13:43:21 oh, I thought you knew that already Mar 28 13:47:39 zmatt: I probably had a nagging thought in the back of my mind, but it didn't surface until I actually tested it :) Mar 28 14:26:11 zmatt: funny. apparently I need to access one register of the CTI after releasing DBGEN it will not clear TRIGOUT Mar 28 14:35:54 hello I am trying to add my own DT overlay I am facing problem - error is bash: line 0: echo: write error: No such file or directory. Mar 28 14:38:08 when I try to add already available DT then it show error : bash: line 0: echo: write error: File exists Mar 28 14:39:58 To add DT overlay I used command like echo "BB-I2C1" > /sys/devices/platform/bone_capemgr/slots and also tried with config-pin overlay BB-UART1 but both won't work Mar 28 14:40:22 can anyone help me ? Mar 28 14:42:50 rishabh_bb_, The slots location moved when the kernel went from 3.x to 4.x. Are you on a 3.x kernel? Mar 28 14:44:54 I am on 4.x , I think I used right slots location because when I cat it then it show output as expected Mar 28 14:46:28 cat slots output 0: PF---- -1 1: PF---- -1 2: PF---- -1 3: PF---- -1 4: P-O-L- 0 Override Board Name,00A0,Override Manuf,cape-universaln Mar 28 14:49:44 The "File exists" error when you try to load it is probably a conflict with the cape-universaln. Mar 28 14:54:11 Dumb question - I'm using the Debian IOT headless image from the website, if I wanted to recompile the kernel to support GadgetFS, what steps would I need to take? Mar 28 14:54:25 I've not really worked with the BeagleBone much before Mar 28 14:55:04 gabe_ : when I try to issue command echo -4 > slots then it shows some message from root then , BBB kick me out of root and when I try to log in again root then BBB stops working Mar 28 14:55:22 Are you trying to disable the universl cape? Mar 28 14:56:01 Yes Mar 28 14:56:05 pen /boot/uEnv.txt and remove "cape_universal=enable" and reboot Mar 28 14:56:07 open* Mar 28 14:57:43 : ok but why when I try to echo name_of_dt > slots_path is not working ? Mar 28 14:58:10 because the universal cape is already using those GPIO pins Mar 28 14:58:25 if it's loaded the pins can't again be repurposed from what I understand Mar 28 14:59:00 ok, let me try first to remove universal cape . Mar 28 15:45:57 hello everybody Mar 28 15:46:09 i have some SERIOUS trouble Mar 28 15:47:34 you see, the kernel 4.4 on the beagle is obviously different than 3.8. (as is expected). so, my collegues got the bad habit on 3.8 to shut down by ripping the cord, right? on 3.8, not a problem. but on 4.4 that causes corruption of the filesystem Mar 28 15:47:44 i verified that via the debug uart on J1 Mar 28 15:47:47 so... Mar 28 15:48:19 any ideas how to "fix" that? cuz i can tell, they wont stop shutting down by unplugging any time soon... Mar 28 16:36:20 hello? anyone there? Mar 28 16:39:59 the easier solution would be to change their bad "shutting down" habits no? Mar 28 16:40:46 well ya know... i am just a peon and i gotta do what my boss asks of me.... Mar 28 16:40:53 so... yeah... Mar 28 16:41:28 nuremberg defense. i am just following orders ;3 Mar 28 16:43:00 i mean, is that a normal thing most users do? i dont see any file system issues doing ordered shutdowns and reboots, so i am guessing its because of that. Mar 28 16:43:23 anyone here an unplugger too? Mar 28 17:07:07 desperate_dude: you just need to make sure your writes are safe and synchronous Mar 28 17:07:25 desperate_dude: nothing beagle specific to it :) Mar 28 17:09:04 so i just set the synchronous in etc/fstab? Mar 28 17:11:05 I'm sure stackoverflow has some ideas on how to do it Mar 28 17:11:47 i know how to edit that, dont worry. Mar 28 17:12:28 well, i mean you probably want more than synchronous Mar 28 17:12:33 you also want a different journal mode Mar 28 17:12:35 something safter Mar 28 17:12:38 safer* Mar 28 17:13:08 the only thing safer than ext4 ordered is ext4 journal isnt it? Mar 28 17:13:20 at least the only i'd know of Mar 28 17:13:23 yeah, data=journall for example Mar 28 17:14:42 basically make sure your FS reverts to a working state after failure Mar 28 17:16:15 i am actually surprised its not already doing that. they had that already covered on kernel 3.8. Mar 28 17:16:33 *shrug* Mar 28 17:16:53 I personally don't care a lot about random downstream forks :) Mar 28 17:19:13 what are you using? Mar 28 17:30:51 I see that will loop block the program if we use hc-sr04-ultrasonic Mar 28 17:46:43 hello Mar 28 17:47:59 i have try to change configure at the kernel with no secsess , and now the bord not boot at all Mar 28 18:10:20 Does anyone here have any experience with GadgetFS on the BeagleBone? I'm able to recompile my kernel with it enabled, however every time I try to mount a device I get an error message about 'gadgetfs' being an unknown mount type Mar 28 19:02:44 Hi! Mar 28 19:03:51 I'am creating the proposal for the Google Summer of Code and the proyect I have in mind is to create a opening system for Makespaces, Hackspaces, Fablabs, etc. The problem is that when the user of this of comunity don't pay, he must not enter in the space. So my solution is to create opening system with NFC, that when the user pass a card, the beagleBone check in the database if the user is in the space. I think that this can be usefu Mar 28 19:04:09 What do you think? Mar 28 19:09:44 desperate_dude: I'm just using openSUSE on my bones Mar 28 19:30:16 yeah zmatt i'm taking a look at the instruction set. i'm starting with cortex-m and moving up. i have fundamental gaps in knowledge though. i actually don't know about the internal architecture beyond whats in athe TRMs Mar 28 19:30:40 so the L1/L2 cache and features are actually beyond me recognizing the vocabulary :-/ Mar 28 19:31:16 i'm also simutaneously learning about gcc... Mar 28 19:31:24 it Mar 28 19:31:31 meh Mar 28 19:45:43 hello Mar 28 19:57:41 honestly though looking at the assembly after compiling gcc makes me question myself. are compilers stupid? Mar 28 19:57:46 or am i missing something Mar 28 19:58:23 because this looks like gcc did a list of `str`s and one `sub` to move arguments onto the stack instead of just using a list of `push` Mar 28 20:01:33 hmm optimization eh Mar 28 20:20:13 Hi all Mar 28 21:46:33 hi, I am new to beaglebone black and I am trying to boot using the microsd without success, the instructions is to press the boot button only right? Mar 28 21:47:46 I have created a file inside the emmc memory to check if I am running the microsd or emmc and it always shows the emmc file I have created Mar 29 01:07:44 ayjay_t: in general, yes compilers are stupid. ;) I can't tell from your description whether this is an instance though Mar 29 01:08:51 ayjay_t: note that dsb even does something on the cortex-m: it drains the write buffer Mar 29 02:54:32 zmatt i just think its doing weird things with the general registers and the stack Mar 29 02:54:55 sometimes it uses push sometimes it uses ldr... Mar 29 02:55:10 it doesnt matter Mar 29 02:55:30 i've seen ldr r0, {r4, #0}; str r0, {r4, #0} Mar 29 02:56:14 even turning on -O1 reduced 25 lines to 4 lines, and i realize now that -O0 literally *means* BE STUPID Mar 29 02:58:06 oh yeah, duh :) Mar 29 02:58:21 I can recommend -Og Mar 29 02:58:28 or -O2 -fno-schedule-insns{,2} Mar 29 02:58:36 i gotta look at man gcc Mar 29 02:58:47 don't know what they do Mar 29 02:58:59 i also need to find a resource for mpu architecture Mar 29 02:59:13 if you let it schedule instructions then you got instructions from different statements interleaved, which is a PITA when single-stepping or when reading disassembly Mar 29 02:59:27 hence the -fno-schedule-insns{,2} Mar 29 02:59:54 (-Og already disables instruction scheduling, but it's sometimes a bit too dumb for my taste) **** ENDING LOGGING AT Wed Mar 29 03:00:03 2017